AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Enhancing Interpretability for Vision Models via Shapley Value Optimization

Kanglong Fan, Yunqiao Yang, Chen Ma

AAAI 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Self-explaining and visual explanation networks · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fine-tune with Shapley value auxiliary lossViT-B · ImageNet classificationAcc@1: 75.0 -> 76.8Improved
fine-tune with Shapley value auxiliary lossResNet-50 · ImageNet classificationAcc@1: 75.3 -> 76.3Improved
fine-tune with Shapley value auxiliary loss (ResNet-50)ResNet-50 · ImageNet classificationFLOPs: 4.1B -> 6.6BHurt
use Shapley self-explaining training instead of post-hoc TokenTMViT-B · ImageNet explanation faithfulnessAOPC: 0.755 -> 0.803Improved
use Shapley self-explaining training instead of post-hoc FastSHAPResNet-50 · ImageNet explanation faithfulnessSaCo: 0.3197 -> 0.5193Improved
use Shapley training instead of B-cosifiedViT-B · ImageNet classificationAcc@1: 75.3 -> 76.8Improved
use Shapley training instead of B-cosifiedResNet-50 · ImageNet classificationAcc@1: 76.5 -> 76.3Hurt
use Shapley training instead of HarsanyiNetcustom CNN · ImageNet-SegmentationmIoU: 66.03 -> 66.96Improved
use Shapley training instead of BagNetcustom CNN · ImageNet-SegmentationmIoU: 66.77 -> 66.38No clear effect
fine-tune EVA-CLIP with Shapley lossEVA-CLIP ViT-B/16 · MS COCO zero-shot retrieval (T2I)R@1: 42.2 -> 42.1No clear effect
fine-tune EVA-CLIP with Shapley lossEVA-CLIP ViT-B/16 · Flickr30k zero-shot retrieval (T2I)R@1: 71.2 -> 70.8No clear effect
use Shapley training instead of B-cosifiedEVA-CLIP ViT-B/16 · MS COCO explanation insertion (T2I)Insertion R@1: 0.1558 -> 0.1679Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy